A reusable instrument to answer one question about any large IPO: is this a good stock from the IPO price? — NOT "is this a good business." The two are different, and conflating them is how the ~86% underperformance base rate eats buyers (Ritter).

Distinct from the S-1 → thesis crosswalk. That instrument (2026-06-02-ai-lab-s1-thesis-crosswalk-playbook) routes a filing's demand-side disclosure into our theses. This one scores the deal itself for the buyer. They compose: crosswalk says "what does this IPO tell us / fund," scorecard says "should anyone buy the stock at this price." Owner: cross-cutting / general (any IPO, not just AI). Derived from the positive base-rate study 2026-06-02-large-ipo-winners-positive-base-rate.

The four factors (each ✓ / partial / ✗)

Decades of IPO winners cluster on four traits; the ~86% that underperform lack them. Ranked by how much each protects the buyer:

1. Digestible entry multiple (price/sales). The dominant separator.

• ✓ ≲ ~10× sales.
• partial ~10–20× — ✓ only if hyper-durable recurring revenue (SaaS) and profitable.
• ✗ > ~20–40× sales.
• Earnings vs sales nuance: rich on earnings but cheap on sales/growth can still pass (Google: P/E 120 but ~10× sales, already very profitable → the multiple could hold/expand). Rich on sales (40–100×) cannot — even 600% forward revenue growth lost to a 90% de-rate (Snowflake).

2. Profitability, or a fast/cheap path to it. Strip BOTH one-time charges and one-time gains before grading.

• ✓ operating-profitable, or a clear near-term path on modest revenue multiples.
• ✗ structural operating losses, or "profit" that is an accounting artifact.
• The symmetry trap: Figma's 2024 GAAP loss was a one-time Adobe stock-comp charge → strip it → still ✓ (really profitable). Cerebras's GAAP profit was a one-time forward-contract gain over a real operating loss → strip it → ✗ (not profitable). Read the operating line and non-GAAP, not the headline.

3. Durable moat / low concentration. The compounding engine — but it only pays off through the multiple.

• ✓ network effects / switching costs / standard-setter, diversified customers.
• ✗ customer or supplier concentration, commodity economics, revocable dependencies (export licenses, single foundry, single customer).

4. Not listed at a frothy peak. Ritter's strongest empirical signal.

• ✓ quiet/down IPO window, or growth fast enough to outrun a top.
• ✗ all-time highs, hot IPO window, or a large day-1 pop.
• Score the post-pop multiple for the public buyer. A +250% day-1 pop converts a digestible entry into an indigestible one — the public buys factor 1 at the peak, not the offer.

The gating insight

Factors 1 and 4 protect the buyer; factors 2 and 3 only pay off through the multiple. A great moat compounds fundamentals, but it cannot outrun a 90×→11× de-rate on any human timescale. So a winner-business (2 ✓ + 3 ✓) can still be a loser-stock if 1 or 4 fail. Always score at TWO entry prices — the IPO offer AND the day-1 close — the gap is the retail trap.

Scoring → verdict

• 4/4 or 3/4 → winner-profile (the rare decade-compounder shape).
• 2/4 → mixed; usually "great business, rich stock" (passes 2+3, fails 1+4).
• 0–1/4 → loser-profile; the 86% bucket / Ritter worst-case ("hot-market, end-of-cycle, high-growth, rich").

This scores expected stock outcome from the IPO/peak price, not the business.

How to run

1. Pull: IPO offer price + valuation; day-1 high/close + that valuation; trailing revenue + growth; GAAP and non-GAAP/operating profit (strip one-offs both ways); customer/supplier concentration; moat; market timing.
2. Compute price/sales at the offer AND at the day-1 close.
3. Grade the four factors ✓/partial/✗; write the verdict.
4. Emit a dated note/investigation; cross-link to the IPO's lane/perspective. For an AI lab, run the S-1 → thesis crosswalk alongside (demand side) — the two instruments answer different questions about the same filing.

Worked examples (validated 2026-06-02)

Post-IPO confirmation: Cerebras −32% from its $350 open (still >$185 offer); Figma −79% from the $115.50 day-1 close and −26% below its $33 offer ten months on. Both round-tripped their buyers — consistent with the 86% base rate.

What the two failure modes teach:

• Cerebras = fails on substance (concentration + thin margin + accounting-flattered profit, at a frothy multiple with no margin for any of it).
• Figma = fails on price/timing (a near-ideal business whose +250% pop and hot window made it a loser stock at the moment retail could buy it).
• SpaceX = fails on both (rich multiple, structural loss, ATH timing; the genuine moat — Starlink — is sold inside a frothy AI wrapper at a frothy price).

Limitations / traps

• Scores the stock from the IPO/peak price, not the business. A 0/4 stock can still be a great company (Cisco/Intel were — buyers at the 2000 peak waited 16–26 years to break even).
• The rare rich-multiple winner needs the NVDA/Amazon move — outgrow the multiple before it de-rates. No catalogued winner did this from >40× sales at a peak.
• Multiples can stay irrational longer than a position can stay solvent — this is a base-rate / position-sizing tool, not a short trigger.
• Day-1 pop data can lag; re-score once the open/close print is final.

Cross-links

• Positive base-rate study (the winners catalog + the four-factor derivation): 2026-06-02-large-ipo-winners-positive-base-rate.
• Companion instrument (demand-side thesis routing): 2026-06-02-ai-lab-s1-thesis-crosswalk-playbook.
• Macro top-signal lens (issuance-surge regime-top): 2026-06-02-ai-mega-ipo-issuance-surge-base-rate.
• One application context (AI mega-IPO lane): 2026-06-02-ai-mega-ipo-fast-entry-liquidity-regime.